Hands-on: studying the long-lasting effects of Tamoxifen treatment on the CNS transcriptome across sexes

Introduction

The past decades have seen large calls for consideration of the sex dimension into research and clinical projects. However as we progress towards more frequent inclusion of this variable, the design of experiments evolved to be much more complex, in particular challenging the analysis steps.

In this hands-on session we would like to focus on a particular transcriptomic dataset where the sex variable was considered in addition to other variables of interest, and illustrate a differential expression analysis aiming at rigorously testing for sex differences and notably the presence of interactions between factors. Following on Frédéric’s presentation, we would like to use this hands-on to draw your attention at power issues potentially affecting conclusions, and how to best report results of your analyses.

Today’s dataset is described in this paper “Tamoxifen induction of Cre recombinase does not cause long-lasting or sexually divergent responses in the CNS epigenome or transcriptome: implications for the design of aging studies”, and focuses on the potential side effects of Tamoxifen treatment, widely used to induce CreERT2 activity in transgenic mouse systems. Tamoxifen acts as an antagonist of estrogen receptor (ER), which could cause differences in response across sexes.

Loading data in R

Read in data

Raw and processed data are available from GEO (accession GSE135752)

Today we will load directly the reprocessed data loaded from the recount3 project, a project which reprocessed uniformly thousands of bulk RNA-seq datasets from human and mouse. See the hands_on_recount3.qmd script where we import and format the DGEList object.

Gene filtering

QC checks

Expression of Xist gene

Xist is a lncRNA regulating the X-chromosome inactivation process in mammals, used to equalize the dosage of X-linked genes between female (XX) and male (XY). It should thus be expressed only in female samples

Expression of Y-chromosome genes

What do you observe? What is going on for some genes? (e.g., Gm20775)

Bigwig files are available through recount3 for some of the samples and allow to visualize the read coverage (for example on the Y-chromosome) on a genome browser. For example you can provide these links to the UCSC genome browser directly:

Principal component analysis

To which principal components contribute which factors? Does this corresponds to the expectations? What do you conclude on the relative effect size of different factors?

Differential expression analysis

Do the number of DE genes correspond to your expectations from the PCA? What would you conclude on the difference of Tamoxifen effects in male and female?

Which genes do you notice here? Try and plot the original logCPM values (uncorrected for tissue effects)

What do you notice? Are some of these genes described in the original paper?

We will now look at the 2 genes reacting differently to Tamoxifen treatment in male and female

How would you describe the behavior of these genes in (well chosen) simple words? Are these genes described in the original paper?

To finish

• “Interpretations of sex-related variation are not always commensurate with the story the data actually tell” (Pape et al. 2024). How do you judge the interpretations of the results in the published paper?

• Discuss potential limitations of the paper and the conclusion about the absence of Tamoxifen effect.

  • What is important to notice in the design of the experiment (tamoxifen treatment duration, batches, number of replicates, …)?
  • Have a look at the results and materials and methods section: which analyses were made and how could they influence the results?

• Limitations of our reanalysis: what could be improved if we had more time?

  • Try subsetting the dataset to one tissue as done by the authors. What do you notice?
  • It is possible that we detect few significant genes for the interaction but that some trends are visible collectively at the gene set level. See the more davanced scripts on the Github repository allowing you to explore this

• “Sex is not a causal mechanism” (Pape et al. 2024): follow-up on some DE genes and try to find which mechanism of action could explain their differential expression patterns.

• Discuss what would be the next steps if that was your research project. Which experiments would you design next? How would you describe the results in a paper?

  • You can also have a look at other related papers (e.g., https://www.nature.com/articles/s41586-022-04686-1, https://www.ahajournals.org/doi/10.1161/ATVBAHA.123.319922)
  • Or look at follow-up papers from the same authors (e.g., https://link.springer.com/article/10.1007/s12035-022-02860-0)

_{Icons taken from http://www.flaticon.com/}